I'm guessing that you don't work on a large compiled application. :)

When you're working with millions lines of C, the speed of the compile decreases (nearly) linearly with the number of cores. Having gobs of memory speeds things up, too, because large portions of your filesystem will be cached in RAM all the time. Throw in a couple of VMs that you use for different build environments and you can easily suck up 4 gigs of RAM.

Depends on what you do. I work in data analysis and our datasets are enormous and since RAM is cheaper than time, we load as much as we can in RAM. 16GB, 6 cores + SSD are standard issue. :-)